EVO-ICL Vault Prediction: A Data Wrangling Framework Integrating Multicenter Big Data and Machine Learning.

Area of Science:

• Ophthalmology
• Medical Informatics
• Machine Learning

Background:

• Accurate prediction of implantable Collamer lens (ICL) vault is crucial for successful refractive surgery outcomes.
• Existing methods for ICL vault prediction may lack precision due to data complexity and variability.
• Multicenter big data presents an opportunity to develop more robust predictive models.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting the ICL vault.
• To assess the utility of a novel data wrangling approach, the digital vault information system (DVIS), in enhancing ML model performance.
• To compare the predictive and classification accuracy of ML models integrated with DVIS against traditional methods.

Main Methods:

• A retrospective study utilized preoperative biometric data from 6715 eyes across five hospitals.
• Mutual information regression identified key predictive parameters.
• A digital vault information system (DVIS) was developed for data wrangling, and ML models (including XGBoost) were trained and validated internally and externally.

Main Results:

• The XGBoost model integrated with DVIS demonstrated superior ICL vault prediction accuracy, achieving low Mean Absolute Error (MAE) of 39.15 μm (internal) and 149.72 μm (external).
• The model achieved an R² value of 0.86 in internal validation.
• For ICL vault classification, XGBoost with DVIS reached 81.4% accuracy (internal) and 57.27% (external), significantly outperforming traditional ML algorithms.

Conclusions:

• The DVIS provides an effective data wrangling strategy, significantly improving ML model efficiency and accuracy for ICL vault prediction.
• This synergistic approach enhances existing ML methods, offering valuable tools for informed clinical decision-making in ICL implantation.
• The developed models show promise for optimizing surgical planning and patient outcomes in refractive surgery.